challenges in refinery optimization –user perspectiveglobal.wonderware.com/en... · fareed khan...

Fareed Khan

Challenges in Refinery Optimization – User Perspective

Simulation for Business Excellence – SE User Conference

6th Oct 2015

Vice President – APC/RTO

Reliance Industries Ltd, India

Disclaimer

Reliance Industries Ltd. © 2015

Any statement, opinion, prediction, comment, or observation made in this

presentation/publication are those of the presenter/author only and in no condition should be

construed necessarily representing the policy and intent of Reliance Industries Ltd. (RIL).

The information presented herein are of the presenter/author’s own and in no way RIL attracts

any liability for any inconsistency or irregularity in terms of the accuracy completeness, veracity, or

truth of the content of the presentation/publication. In addition, RIL shall not be liable for any

copyright infringement and misrepresentation for the presented content as the content is

presumed in good faith to be a creation of presenter’s/author’s own mind.

The scope of this presentation/publication is strictly for knowledge sharing purposes and not

necessarily to provide any advice or recommendation to the audience/readers. Any endorsement,

recommendation, suggestion, or advice made by the presenter/author shall be in his personal

capacity and not in professional capacity as an employee of RIL. Any person acting on such

endorsement, recommendation, suggestion, or advice will himself/herself be responsible for any

injury/damage.

2

� Reliance – at a lance

� Optimize in Real Time - Preamble

� Brief Introduction to RTO

� Reliance Experience & Initiatives

Content

3

Reliance – At a Glance

Reliance Industries Ltd. © 2015 4Annual Report – 2014-15

Largest Refinery complex globally at a single Location– 1.4 MMBPD of Crude

Processing

Challenge: Optimize in Real Time – Preamble

� Agility

� Create system to capture opportunities as they arise

� Streamline People process

� Provide in time correct economic information

� Model Accuracy

� Unit RTO Models need to mirror the real plant

� Reactor models need be fast, robust, and complex enough to capture economics with mathematics amenable to easy

optimization

� APC models need acceptable fidelity

� Inferential needs wide range applicability

5Reliance Industries Ltd. © 2015

Challenge: Optimize in Real Time – Preamble

� Integration

� Planning and Scheduling system integration with APC/RTO

� Blending System integration with Scheduler, RTO and APC

� Orchestring a complete refinery wide optimization and control solution

� Expertise

� Shortage of good engineers resulting in sub optimal solutions, not realizing full economic potential.

(Knowledge of operation, technology, engineering, DCS, instruments, economics & vendor tools are required)

� Technical Services – Lack of taste for optimization and open equation software


Available Systems – Preamble

� Advanced process control (APC)

� APC software -- Mature and good record of success

� Inferential software -- Accuracy issue

� Real Time Optimization (RTO) systems

� RTOs -- Varied degree of success

� Integration to APC -- Minor customizations required

� RTO Analysis -- Not mature

� RTO Reporting -- Needs customization

� Planning & Scheduling System

� Planning Software -- Difficult configuration, no GUI & Stream properties

� Scheduling Software -- Runs un-constrained, difficulty in modelling all aspects of operation & people process

� Blenders -- Analyzer Reliably issue


Introduction: Control Hierarchy

Scheduling

Optimization

LP - QP

APC APC

DCS

Empirical & Dynamic model

DMCplus

Rigorous, Plant wide & Steady

state

ROMeo

Simple, Site wide & Steady

state

msec

min

hrs

Weeks &

Months

Instrument & Plant

Multiple or single loop no Model

PlanningSimple, Enterprise wide &

Steady state

Days


APC - Limitations

9

Feed Rate

ProfitConstraints-Safety-Equipment -Quality

Optimum at Constraint

Features: � Dynamic Plant Model� Regressed Model from plant data� Optimum always at constraints� Linear Objective & Constraints (LP or QP)

Limitations: � Linear Gain Models� Linear Cost Function� Limited First Principle Modelling� Optimum always at constraints

Reliance Industries Ltd. © 2015

RTO – For Floating Optimum

Reactor Temperature

Profit RTO

Floating Optima

Constraints-Safety-Equipment -Quality

�Fundamental Model from first principles�Steady State Plant Model�Optimum floating�Nonlinear Objective & Constraints (QP)�Needs Steady State Detection

APC


Types of Simulation – SM

Sequential Mode Solvers (SM) – Aspen Plus, Petro-SIM, PRO II, UNISIM, HYSYS

Feed

Product

�Give inputs �Calculate outputs

�Initialization is easy�Can solve differential equations for Reactors �Solves one by one and proceeds�Needs different solving algorithm for each equipment�Difficult to solve for recycle streams and heat integration�More difficult to optimize �Still more difficult to respect constraints


Types of Simulation - EO

Equation Oriented Solvers (EO) – Invensys Opera, Aspen DMO, Honeywell NOVA

Feed

Product

�Give outputs�Calculate inputs

�One solving algorithm for complete plant flowsheet�Easy for recycle streams and heat integration�Easy optimization and constraint handling�Difficult to initialize�Needs equation and variable information�Needs special reactor model consideration�Solves only algebraic equations


What is SM & EO

SM - Closed Equations

Q = U *A *LMTD

U = Q / (A * LMTD)

LMTD = Q / (U *A)

dF/dV = -ri

Features:

� Here 3 formulation for one equation�Would needs 3 different solving algorithm�Only Simulation

EO - Open Equations

0 = Q - U *A *LMTD

Features:

�One formulation for one equation�Same equation for : Simulation Parameterization, Data Reconciliation & Optimization�Needs variable information (const & Calc)


Additional Requirement for EO

Variable Definition (Const & Calc):

0 =3 +4*X1 +5*X2 +7*X3

0 =2 +X2^2 -2*X4

0 =4 +9*X1 +3*X4

0 =6 +5*X1 +4*X3

1 1 1 0

0 1 0 1

1 0 0 1

1 0 1 0

Sparsity Matrix

Jacobian Matrix

4 5 7 0

0 2X 0 -2

9 0 0 3

5 0 4 0

Input variables needs to be defined as constants

Output variables needs to be defined as calculated

Variables location in an equation needs to be defined in a sparsity

pattern. Only 2% or less non zeros are there in process plants.

Sparse Matrix Algorithms Reduce memory requirement and

computational time

Direction required to solve the equations is defined in a Jacobian

matrix. These are slopes or partial derivatives, generally determined

analytically

A second order numerical derivative information is defined in a

Hassian matrix that is calculated from derivative information


Additional Requirement for Reactors in EO

� Reactor Models are Differential Equations

� EO Solves only Algebraic Models

� Differential Eqs. are converted to algebraic Eqs. by collocation

Example:

y = ln(x)

dy/dx = 1/x

at x=1, y = 0

Assumes a solution equation as:

y = a+bx+cx^2

dy/dx = b+2cx

at x = 1, dy/dx =1

& x = 1.1, dy/dx = 1/1.1

y = -1.4545 +1.909*x -0.4545*x^2 Spline 1 Collocation Spline 2


Automatic RTO Sequence

Measurement Validation

Plant Steady?

Implement Optimization points (Ramper) Line-Out period 60 minutes

Optimization Calculation

Data Reconciliation / Parameter Calculation

Plant Steady? Wait one Minute

Less than 1 minute

15 minute

10 minute

Measurements

Economic Values

Controller Limits

Control and opt.

Status Info.

Selected Plant

Measurements

Selected Plant

Measurements and

Controller Limits

No

No


RTO Solver Algorithm

� Sequential Quadratic Programming

1. Given Xk , Set k = 0

2. Evaluate Functions and derivatives

3. Compute search direction, Delta Xk

4. Check for convergence

5. Compute step length, ak6. Set Xk+1 = Xk+ ak Delta Xk

k = k+1

Go to step (2)

Sort of Newton Raphson with variable step size


RTO Additional Considerations

1. Normalization of Variables to account for Units of Measure

2. Normalization of Equations to avoid difference in numerical values

3. Smoothing thermodynamic discontinuities (Complementarity)

4. SM Initialization Model for filling initial values

5. Configuration of Parameter case, Reconciliation case, Simulation case

and Optimization case.

6. Configuration of Objective Function (Merit Function)

7. Configuration of Online sequence logic

8. APC, Data base, DCS and Users connectivity

9. Pricing and Quality information gathering and report generation

10. Steady State Detection


� RTO is used to capture a floating optima that is not easy to calculate

� Recycle and heat integration in Closed Equation model is cumbersome and on many occasions

impossible to use for optimization purposes

� Open equation models needs good initialization and special consideration for Reactors

� RTOs needs current Prices, feed quality information and requires more attendance

RTO Recap


Initiatives @ RIL

APC initiative� STAR Corporate License in 1994

� SMCA PMD License in 1994

� DMCplus HMD Cracker unit License in 1998

� DMCplus Corporate License in 2001

RTO initiative� NOVA RTO Corporate License in 1997

� Aspen RTO unit License for HMD Cracker 1999

� Aspen RTO unit License for JMD FCC 2006

� ROMeo RTO Corporate license from Invensys 2012

Current Activities� In-house implementation group for both APC and RTOs� More than 800 pages APC/RTO manual written to standardize procedures� Use of First principle based inferential for use in APC� Development of reactor models for use in RTO� Monitoring KPIs and Dashboard� Expanding use of Open Equation based models for debottlenecking and troubleshooting� Develop integrated methodology for refinery wide control


Our Experience

� APCs for Operator entry and control at Refinery wide level – Refinery Co-ordinating Control (RCC), Refinery Co-ordinating Optimiser (RCO)

� Graceful system degradation at each level, (DCS, APCs, RTO, RWO)

� Simple enough so that it is amenable to human understanding


Simplified Mass and

energy Balance

Quality Constraint Handling

Sectional Constraint

Handling

Elaborate Mass and

Energy Balance

Refinery-wide Coordinating Control (RCC)

Unit APC

DCS

SP Constraints

Unit RTO

ConstraintsET

Unit APC

Unit RTO

Unit APC

Unit RTO

Refinery-wide Simplified Optimizer (RCO)

QualityET

Plant Constraints

Our Experience

� Start from product dispatches and product specs and work back



� Aim to build one detailed model of each unit to rely on if the RWO is simplification

MTF Tanks CrudeVGOHT

FCC Reformer

Assay

BlendsBlend

Adjuster

Crude RTO FCC RTO

Composition Transfer Composition Transfer

REF RTO

Through Database

Direct

Direct

NHTR

Our Experience

Current Focus


� RWO model to solve within reasonable time by using

� Simplified Reactor

� Simplified Column models

� Less rigorous energy balance

� Reduced constraint set

� Component balance &

� Stream properties calculation


Scheduling

Refinery-wide Coordinating Control (RCC)

Unit APC

DCS

LP Planning & Scheduling

Narrow Assay Applicability

SP Constraints

Unit RTOConstraintsET

Unit APC

Unit RTO

Unit APC

Unit RTO

Refinery-wide Simplified Optimizer (RCO)

QualityET

QP Embedded in Scheduling

Plant Constraints

Feed Blend Optimizer Product Blend Optimizer

Long Term Planning & Scheduling

Wide Assay Applicability for Buying

Manual Base + Delta Vector

Automated Base + Delta Vector

LP Embedded in Scheduling

Constraint

Violation

Closed Loop Header Property Control

• Refinery Wide Optimizer (RCO) to guide APCs

• Blending & Scheduling system to Guide RCO

• Independent Layer for human analysis

• Graceful system degradation with each layer

Optimization & Control Scheme

Centralized RTO Setup @ Reliance

� All Configuration work by Reliance Engineers

� Consulting services from Invensys & KBC for initial projects

� Pool of servers at Central Location

� User login from all Sites

� All project work at one place

� Initial commissioning backup

Mumbai Users


Monitoring KPI and Dashboard


Summary


� Look at these systems as repository of knowledge and company's learning

� Create in-house implementation group

� Expand the use of models to plant debottlenecking and troubleshooting studies

� Involve operations, technical services, schedulers and planners, it indicates the success level of installed

system

� Involve top management for Focus, attention, and visibility as objectives of stake holders are varied and

sometimes clash with each other because of safety, economics, reliability and comfort level

challenges in refinery optimization –user perspectiveglobal.wonderware.com/en... · fareed khan...

Documents